Research On Target Detection In Synthetic Aperture Radar Image

Synthetic Aperture Radar (SAR) is an advanced observer which works on microwave band. Compared to optical and infrared imaging system, SAR has many advantages such as having high resolution, having ability of imaging large area, less weather and time limitations, also having the ability of penetrating through clouds, frog and grand plants. With these advantages, SAR has been widely used in both military and civil applications.Through systematically theoretical and experimental research, and based on SAR image's inherent properties, we build and improve SAR target automatic detection and contour extraction methods. First, we study the statistical models that are suitable for describing SAR data acquired from metallic target, land clutter or the mixture region of target and land clutter. Based on the results, we propose a mixture statistical model which can better describe the randomness of SAR data from the mixture region of land clutter and metallic target. The parameter estimation methods are given, and goodness of fit tests are conducted.For the target detection, we study the traditional Constant False Alarm Ratio (CFAR) detection algorithm, and improve the algorithm by modifying the clutter selecting method. The experimental results have shown that the improved CFAR algorithm has better performance than the traditional one in terms of better target pixels retaining ability. CFAR algorithm does not consider the statistical properties of target in SAR image, so it is a suboptimal detection method. In order to get a better performance in detection and bring the target information into detection algorithm, we also propose a Generalized Likelihood Ratio Test (GLRT) algorithm based on K-Lognormal mixture model, and the results show the GLRT algorithm has better performance.Finally, for the edge detection, we first briefly introduce Snake's fundamental theory and its numerical solution. And then, aim at the problems of speckle noise and blur edge in SAR image, we have made several modifications for the Snake model, which makes the Snake model method can be usefully applied for SAR target edge detection and contour extraction.